I've also seen the same issue. My hosts are using the HP custom 5.5 image too. I have an existing ticket open with VMware on the issue, but they don't have a resolution at the moment and at the time hadn't run into it before.
In my case, I typically see the "malformed response" error more than the "terminated by server" one, but both have occurred. Connecting directly to the host doesn't allow the console to open either. Workarounds to date have been to vmotion to a working host, or reboot the host. I haven't seen it occur on all of my hosts.. so far it's only occurred on one of my clusters.
However, I have another issue that may (or may not) be related to the HP image, which occurs after a random amount of time where the host effectively is out of resources to spawn new processes, which affects certain tasks. An easy way to identify that situation is that SSH drops the connection prior to being able to enter a password and you'll see errors in the syslog complaining that cron can't fork due to lack of space (but the server isn't out of space).
Thanks
-Joshua
